模糊图像复原的高阶全变差正则化模型构建  

High-order Total Variation Regularization Model for Image Deblurring

作　　者：余瑞艳[1] 

机构地区：[1]长江大学一年级教学工作部,湖北荆州434025

出　　处：《长江大学学报（自科版）（上旬）》2014年第3期40-43,3,共4页JOURNAL OF YANGTZE UNIVERSITY (NATURAL SCIENCE EDITION) SCI ＆ ENG

基　　金：国家自然科学基金项目(11201039;61273179);湖北省教育厅重点项目(D20101304)

摘　　要：模糊图像复原是一类典型的不适定反问题。为保证复原过程的稳定性,在二阶梯度域构建了模糊图像复原的高阶全变差正则化模型。针对该模型数值求解的困难,将原问题分解成Tikhonov反卷积和高阶全变差图像去噪2个子问题,并分别对其设计高效的数值求解格式。试验结果表明,该图像复原方法在快速恢复图像细节信息时能够有效地抑制振铃效应和阶梯效应的产生。Image deblurring is a typical ill-posed inverse problem in practice.To guarantee the deblurring quality,a highorder total variation(HOTV)regularization model is proposed in this paper.However,the proposed model usually suffers from numerical inefficiency.In order to achieve an efficient numerical solution,the original deblurring model is decomposed into two subproblems,i.e.,Tikhonov deconvolution and HOTV-based image denoising.The corresponding efficient solution algorithms are proposed to solve the subproblems.The experimental results demonstrate that the proposed deblurring method could achieve an efficient deblurring result while suppressing both ringing and staircase artifacts.

关 键 词：图像复原 高阶全变差正则化模型 Tikhonov反卷积 高阶全变差图像去噪 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]